From Jonathan Simon of the Election Defense Alliance:
Mark, just a quick gloss on this:
“As the Election Defense Alliance showed in a meticulous analysis, Brown ‘won’ handily (by over 5%) wherever votes were counted electronically—but, in those few districts where the ballots were hand-counted in the open, Scott Brown lost to Martha Coakley (by almost 3%).”
Although the number of ballots counted by hand was just a couple of percent of the total in the state, there were actually a lot more than a “few” districts (71 out of a total of 351) where that hand-counting took place. The phrase “those few districts” could leave the impression that the hand-count/machine-count disparity derived from only a tiny subset of voters. In fact, the hand-count “sample” was not only huge in statistical terms, but also fairly extensive geographically (i.e., not just a couple of weird enclaves).
The real statistical point of our analysis was that the hand-count voters were to the right of what would have been a random, representative sample. If the sample had been representative, the odds against an 8% marginal spread between hand-counts and opscan counts were truly astronomical (literally quadrillions to one); and the “sample” was to the right of that, which only increased those odds. That is as close to numerical “proof” as it gets.